Are Magnitudes from Sparse Networks Biased by Unequal Azimuthal Distribution of Stations?

Abstract:

Earthquake magnitudes are normally an average of magnitudes calculated at individual stations, which rarely yield identical values. For dense networks, azimuthal variations related to radiation pattern or path effects are likely “averaged out”. This may not be the case for sparse networks with non-uniform station distribution where oversampling at one azimuth with respect to another may produce a biased magnitude. Using a dataset of 1745 eastern Canadian earthquakes with magnitudes calculated at a minimum of ten stations, the effect of unequal station distribution on magnitudes was explored. A simple comparison was made averaging station magnitudes for predefined azimuthal windows and then taking the mean of these values to be the event magnitude. These were compared to magnitudes derived via the current practice of treating all stations equally. The mean difference was 0.00. There was no correlation between the magnitude difference and either the total number of populated azimuthal windows or windows containing only a single station. Outlier events were examined manually. Almost all were events with a large number of stations on soft soil sites in southern Ontario. The southern Ontario stations cover a narrow range of azimuths for most eastern Canadian earthquakes and therefore fall within a small number of azimuthal windows. A second test treated stations in the Charlevoix Seismic Zone (CSZ) where station density is high as a single point and all other stations individually. The mean difference was 0.00 with respect to the original magnitudes and there was as much variation among stations within the CSZ as for the non-CSZ stations. For eastern Canada, there is no obvious benefit derived by excluding stations from magnitude calculations in an attempt to reduce azimuthal bias. The effect of site corrections, while small, is more significant. Whether these observations are applicable globally is to be determined.